Pretrained deep net (Vgg-F)

I have downloaded a pretrained model Vgg-F from the site (https://www.vlfeat.org/matconvnet/pretrained/)

When i try to pass my image without training i got the following error

net = load('imagenet-vgg-f.mat') ;
layer ='fc8'
feat = activations(net,img,layer,'OutputAs','columns');

Check for missing argument or incorrect argument data type in call to function 'activations'.

Error in (line 81)

feat = activations(net,img,layer,'OutputAs','columns');

Please how to solve the problem?

Answer 1

[colordepth]

The error you're encountering is because the network loaded is a struct, not a "SeriesNetwork" or "DAGNetwork" object, which the "activations" function requires. You can read more about the function by using the following command:

doc seriesnetwork.activations

To work with this network, please refer to the documentation for the MatConvNet toolbox at https://www.mathworks.com/matlabcentral/fileexchange/47811-vlfeat-matconvnet. The functions you need are "vl_simplenn_tidy" for initializing the neural network and "vl_simplenn" for running it on an image. These functions are provided in the toolbox.